The objective of this study was to test whether non-invasive assessment of respiratory stroke volume variation (ΔrespSV) by oesophageal Doppler monitoring (ODM) can predict fluid responsiveness during surgery in a mixed population. The predictive value of ΔrespSV was evaluated using a grey zone approach.Ninety patients monitored using ODM who required i.v. fluids to expand their circulating volume during surgery under general anaesthesia were studied. Patients with a preoperative arrhythmia, right ventricular failure, frequent ectopic beats, or breathing spontaneously were excluded. Haemodynamic variables and oesophageal Doppler indices [peak velocity (PV), stroke volume (SV), corrected flow time (FTc), cardiac output (CO), ΔrespSV, and respiratory variation of PV (ΔrespPV)] were measured before and after fluid expansion. Responders were defined by a >15% increase in SV after infusion of 500 ml crystalloid solution.SV was increased by ≥15% after 500 ml crystalloid infusion in 53 (59%) of the 90 patients. ΔrespSV predicted fluid responsiveness with an area under the receiver-operating characteristic (AUC) curve of 0.91 [95% confidence interval (95% CI): 0.85-0.97, P<0.0001]. The optimal ΔrespSV cut-off was 14.4% (95% CI: 14.3-14.5%). The grey zone approach identified 12 patients (14%) with a range of ΔrespSV values between 14% and 15%. FTc was not predictive of fluid responsiveness (AUC 0.49, 95% CI: 0.37-0.62, P=0.84).ΔrespSV predicted fluid responsiveness accurately during surgery over a ΔrespSV range between 14% and 15%. In contrast, FTc did not predict fluid responsiveness.